search

mks0601 / TF-SimpleHumanPose

TensorFlow implementation of "Simple Baselines for Human Pose Estimation and Tracking", ECCV 2018
335 stars 90 forks source link

Real time model #61

Closed MozhganPourKeshavarz closed 4 years ago

MozhganPourKeshavarz commented 4 years ago

I thank you so much for your contribution, I'm going to train a human pose estimation model on my own dataset and use it in browser (server-less approach), at the first step, I just want to test a pre-trained model, If I use this repo, I would grab a human detection model and then fed its result to your pre-trained model, is it a good idea? do you think it can be work in real time with reasonable accuracy? i would be appreciated if you could give me a hint, thanks.

mks0601 commented 4 years ago

Depends on which human detection model will you use, but I think it would be hard because this model is not lightweight for the purpose of server-less environment. You should search some mobilenet things

thiranjayamunasinghe commented 2 years ago

@MozhganPourKeshavarz Hey, Did you try this approach in a serverless environment. If so, did you get reasonable accuracy?